Search papers, labs, and topics across Lattice.
The authors introduce MolCryst-MLIPs, an open database of MACE-based machine-learned interatomic potentials (MLIPs) for nine molecular crystal systems, fine-tuned from the MACE-MH-1 foundation model using an automated machine learning pipeline (AMLP). Fine-tuning achieves a mean energy MAE of 0.141 kJ/mol/atom and a mean force MAE of 0.648 kJ/mol/Angstrom across all systems, demonstrating high accuracy. Molecular dynamics simulations validate the models' dynamical stability and structural integrity, making them suitable for production MD simulations.
Pre-trained MLIPs for molecular crystals are here, offering immediate access to accurate and validated potentials for simulating complex material behavior.
We present an open Molecular Crystal (MC) database of Machine-Learned Interatomic Potentials (MLIP) called MolCryst-MLIPs. The first release comprises fine-tuned MACE models for nine molecular crystal systems -- Benzamide, Benzoic acid, Coumarin, Durene, Isonicotinamide, Niacinamide, Nicotinamide, Pyrazinamide, and Resorcinol -- developed using the Automated Machine Learning Pipeline (AMLP), which streamlines the entire MLIP development workflow, from reference data generation to model training and validation, into a reproducible and user-friendly pipeline. Models are fine-tuned from the MACE-MH-1 foundation model (omol head), yielding a mean energy MAE of 0.141 kJ/mol/atom and a mean force MAE of 0.648 kJ/mol/Angstrom across all systems. Dynamical stability and structural integrity, as assessed through energy conservation, P2 orientational order parameters, and radial distribution functions, are evaluated using molecular dynamics simulations. The released models and datasets constitute a growing open database of validated MLIPs, ready for production MD simulations of molecular crystal polymorphism under different thermodynamic conditions.
